LED products have started to displace traditional light sources in many lighting applications. One of the commonly
claimed benefits for LED lighting products is their long useful lifetime in applications. Today there are many
replacement lamp products using LEDs in the marketplace. Typically, lifetime claims of these replacement lamps are in
the 25,000-hour range. According to current industry practice, the time for the LED light output to reach the 70% value
is estimated according to IESNA LM-80 and TM-21 procedures and the resulting value is reported as the whole system
life. LED products generally experience different thermal environments and switching (on-off cycling) patterns when
used in applications. Current industry test methods often do not produce accurate lifetime estimates for LED systems
because only one component of the system, namely the LED, is tested under a continuous-on burning condition without
switching on and off, and because they estimate for only one failure type, lumen depreciation. The objective of the study
presented in this manuscript was to develop a test method that could help predict LED system life in any application by
testing the whole LED system, including on-off power cycling with sufficient dwell time, and considering both failure
types, catastrophic and parametric.
The study results showed for the LED A-lamps tested in this study, both failure types, catastrophic and parametric, exist.
The on-off cycling encourages catastrophic failure, and maximum operating temperature influences the lumen
depreciation rate and parametric failure time. It was also clear that LED system life is negatively affected by on-off
switching, contrary to commonly held belief. In addition, the study results showed that most of the LED systems failed
catastrophically much ahead of the LED light output reaching the 70% value. This emphasizes the fact that life testing of
LED systems must consider catastrophic failure in addition to lumen depreciation, and the shorter of the two failure
modes must be selected as the system life. The results of this study show a shorter time test procedure can be developed
to accurately predict LED system life in any application by knowing the LED temperature and the switching cycle.